**Disclaimer** This post is barely a drop in the bucket when it comes to the concerns and impact around AI. Even the nitty gritty issues are loaded with nuance. Even "what is AI?" is a terrible question to try to answer. I've tried to keep the language layman friendly. Also, there's plenty I don't know or understand and my opinions are ever-changing.
AI is an avalanche coming down the mountain of humanity.
How do you react when you look up the mountain?
Here at GM Assistant, we've drawn a line in the sand of what we find acceptable. If you don't agree with where we drew the line, that's perfectly fine! Below, I will muse around some concerns and our philosophy on them. Our focus is to be transparent so that users and prospective users know where we stand, it is not meant to change your mind. In fact, please do your own research.
Concern: Unauthorized use of intellectual property in the pre-training and/or fine-tuning of AI models.
We are trying to strike a tricky balance here. We do not generate AI-created artwork, writing, or creative content. Our AI is strictly used for organizing, transcribing, and analyzing user-provided input; not for producing anything new.
Because we don't generate anything, we can't plagiarize or replace human artists, writers, or creators. Even if the AI models we use were trained on a mix of data, including copyrighted works, we are not using them in a way that replicates or profits from creative expression.
Our focus is on assisting human creativity, not replacing it. A GM's world, story, and characters are their own; our AI simply helps them track and organize those ideas more efficiently. No one's livelihood or artistic integrity should be impacted by our tool.
Concern: Use of AI replacing work that could potentially be performed (or performed better) by a human.
Unfortunately, the same AI technology that's curing disease is simply being used when it shouldn't be. When you have a hammer, everything looks like a nail. I'm looking at you, AI written emails. And a dishonorable mention to 99% of Facebook posts while we are at it.
We don't use AI for things such as generating advertisements, spamming email lists, writing blog posts, etc. When it comes to what we produce for our product, I really want everything to be deliberate and well crafted.
Concern: A plethora of half-baked, sub-par AI tools and content flooding the market.
There's already an overwhelming amount of AI-generated content and AI tools that don't solve actual problems.
We are addressing this at GM Assistant by ensuring we are solving real problems and providing significant value. Not just calling some API on your behalf and up-charging you. Which is scummy and not a sustainable business approach anyways.
Concern: The massive energy needs to train AI models, and to a much lesser extent, use them.
While I actually believe/hope the big money behind AI will eventually get us over the hump in solving our energy crisis, in the short term there are real implications. Compute in data centers is straining our power plants, burning fossil fuels, and requiring copious amounts of water pumped from local water sources to cool said power plants (water is evaporated via cooling towers).
The big power consumption of AI happens in the AI model training process, which is a one-time energy usage. Inference is using the model after it's trained, while still requiring power, it's way less than training. They say inference for a ChatGPT "message" takes 10x the power of a pre-AI Google search.
We run most of our AI inference on my old gaming computer under my desk, powered by my whole house solar. We don't do any AI training.
I'd say power-wise, for every hour of audio we analyze for our users, it's like I played a video game with modern graphics for about 5 minutes. When we need to scale, we cloud-burst to Google Cloud. Whom I assure you, is doing that inference as power efficiently as they can. To do the math, hypothetically, 1000 users analyzing two, 4-hour sessions a month is as if I played video games for 27.8 days of that month. I.E. half as much power as my AC unit in the summer months.
---
With all this said, I'm in the same pickle jar as everyone else. If I knew how to get these technologies regulated or ablated from our society, I'd be all for it (another loaded topic). Heck, let's get rid of the internet while we are at it — PLEASE. But alas, I am forced to adapt, choosing to attempt to out-ski the AI avalanche to avoid being buried in it.
- Alex